Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
An Agenda for Future Research BACKGROUND As the previous chapters of this report indicate, the past decacle has witnessed considerable progress in efforts to mea- sure the basic dimensions of criminal ca- reers ant! the factors associated with them. Much of that progress is attribut- able to a growing interest in the criminal career paradigm as an appropriate and effective way to study the people proc- essed by the criminal justice system, their patterns of offending, the factors that in- fluence their offending, and the potential influence of public policies both within the criminal justice system and outside it-on indiviclual criminal careers. The earlier chapters have also shown that the state of knowleclge regarding criminal careers is increasing rapidly. Some basic estimates of the mean values of career dimensions among different subpopulations have begun to be clevel- opecI, and some estimates have been found to be consistent in independent studies with different modes of measure- ment, lending greater confidence to the results. Estimates of the individual of fending rate, A, have been particularly info`~ative because they highlight the considerable diversity among offenders. The emerging picture of career cliver- sity has motivated efforts to develop pre- dictors of those career dimensions. Such attempts have been stimulating and pro- vocative for both researchers and deci- sion makers, ant] have also encouraged critical attention to the role that predic- tion in its current state of development should have in criminal justice decision making. That attention has reinforced the necessity for developing better informa- tion on the strength and validity of vari- ous candidate predictors and has lecl to a focus on the underlying technical issues involved in classification and prediction. In addition, attention to the ethical ques- tions that are involvecl has resulted in some convergence of the different view- points. The policy value of current and future research on criminal careers is already substantial. This value derives partly from a recognition that an indiviclual of- fender the unit underlying the criminal career paradigm- is the primary subject I98
AN AGENDA FOR FUTURE RESEARCH of decision making within the criminal justice system. For example, identifying life events or individual characteristics that influence the onset of criminal activ- ity and determining the effect of preven I99 elements of this high-priority project, we turn to some important specific research questions that should be pursued; many of them could be addressed within the context of the proposed longitudinal five interventions in reducing the likeli- study, but they need not be. In all cases, honcl of filcher`? criminal h~h~vior urn they derive from concerns that have been raiser! in earlier chapters of this report about important issues that need more careful measurement or validation or that are not known at all. These research ques tions are organized into three major cate gories: (1) exploration of promising inter vention strategies; (2) continuation of basic research on criminal career dimen sions and their correlates; and (3) new directions in measurement and mocleling of criminal careers. The final section addresses some organizational arrange ments for carrying out the proposed re search agenda as well as for facilitating effective and responsible implementation as well-founded research results emerge. ~ of future criminal behavior are important priorities for research with ma- jor policy implications. In a similar vein, the available research indicates that many youths who have some contact with the juvenile justice system do not persist in criminal activity as adults, but further research is needed on the factors that influence the termination of a criminal career. Basic research on the nature and measurement of criminal careers, and policy research on the effect of various intervention strategies at different stages of a criminal career, wfl! contribute to the development of better policy options. The criminal career paradigm high- lights the need for longitudinal research on both offenders and nonoffenders. And the potential for results from research is enhanced considerably by the opportu- nity to capitalize on methodological de- velopments in other fields that are exam- ining the longitudinal progression of "careers" in such areas as substance abuse, mental health, and employment. Building on developments over the past decade, our research agenda first presents a major new research initiative that is now warranted as the next step-a prospective longitudinal research project. Such an effort would provide a basis for tracking individual criminal careers over time and for linking the characteristics of an incliviclual's career to other life events and experiences, especially to interac- tions with the criminal justice system. A longitudinal study couIcI now build from the collection of measurements ofthe key dimensions of the criminal career to link them to factors that can more properly be viewer! as causal. Following a discussion of some of the A LONGITUDINAL STUDY OF CRIMINAL CAREERS Many issues about criminal careers cannot be adequately addressed in cross- sectional research: the influence of vari- ous life events on an inclividual's criminal career; the effects of interventions on ca- reer development; and distinguishing between developmental sequences and heterogeneity across individuals in ex- plaining apparent career evolution. An- swering these and relatecI questions re- quires a prospective longitudinal study of individuals of different ages. Recognition of the inherently Tongitu- dinal and dynamic characteristics of crim- inal careers has led over the past several years to a number of studies that involved longitudinal data collections. While each of these studies has been valuable in its own way, they have been limited, primar- ily because of the small samples of seri- ous offending that are involved. They
200 have provicled important theoretical in- sights into the development of criminal careers and methodological insights into measurement of the phenomena; hence, they have prepared the way for undertak- ing a major longitudinal study of offend- ing. Even though the fundamental criminal career phenomenon of interest is inher- ently longitudinal, studying it Tongitudi- nally rather than in a sequence of cross- sectional samples is a matter of design choice, for both economic and scientific reasons. First, drawing and following a single longitudinal sample may well be less expensive than assembling the nec- essary records on a comparable sequence of multiple cross-sectional panels. Also, a single longitudinal pane! avoids the vari- ation across individuals in either a single cross-sectional sample or in a sequence of cross-sectional waves. Thus, a longitudi- nal design is probably more efficient in terms of cost per unit reduction of error. Second, the recovery of life history events for an individual is subject both to considerable memory decay and to errors in the sequencing of events in time. To describe and understand criminal ca- reers, it is essential to recover as many offenses and other life events as possible and to order them precisely in time. In repeated cross-sections with different in- dividuals, information would have to be collected for a very long time on any individual to attain the necessary level of accuracy. Repeater! cross-sections involv- ing the same individuals are needed to achieve that level of accuracy. Further- more, organizing information about in- divicluals' group involvements requires reconstruction of their social networks, and a prospective longitudinal design will more readily permit the recovery of information on those relationships. Thus, the pane! concludes that future research on criminal careers should in- clude, as a major component, a prospec CRIMINAL CAREERS AND CAREER CRIMINALS five longitudinal study with several co- horts. Such a study would offer an arena within which to pursue a broad range of issues related to criminal careers, such as the developmental experiences engen- dering compliant behavior, the behav- ioral precursors of subsequent delin- quency and criminality, the influence on subsequent behavior of interactions win the juvenile and criminal justice systems, and the factors associated with career ter- mination. Generally, the project would seek to make detailed measurement of the initiation and termination of individ- ual criminal careers, including a focus on the distinction in those patterns among different kinds of crime, especially be- tween the more and the less serious. With those criminal career observations, as well as information on important life events (e.g., family conflict, divorce, school failure or dropout, drug initiation, processing by the criminal justice system, employment, marriage, etc.), the sequen- tial ordering of events will begin to sug- gest directions of respective influence. This research would properly be aug- mented by ethnographic studies to de- velop information about group and com- munity influences that might not be elicited in the individual reports. Longi- tudinal projects similar to what is pro- posed here have had quite impressive results. The panel believes that a significant effort should be invested now in the de- sign of a longitudinal study and that the federal government should take the lead in initiating what promises to be a na- tional resource for researchers as well as for criminal justice policy makers. The initial design effort should include (1) defining the sampling frames and speci- fying the appropriate sample sizes in var- ious strata; (2) developing Me sampling strategy, including making the selection between representative Copulations 1 high-risk ~"la~vllo VllG111148, ·llV1 ~J.~1- aria populations offering more in
AN AGENDA FOR FUTURE RESEARCH stances of serious offending and choosing between drawing local samples for con- venience, efficiency, and record com- pleteness, and national samples for breadth of representation; (3) cleveloping reinterview schecluTes that are frequent enough to capture event sequences accu- rately yet not frequent enough to influ- ence behavior; (4) formulating and testing appropriate instruments; ant! (5) identify- ing and planning experimental interven- tions with randomly drawn subsamples. Any such longitudinal project should have a breadth of perspective that repre- sents the diversity of disciplinary and theoretical orientations that are relevant to research on criminal careers. In adcli- tion to the disciplines concerned directly with criminal careers, those perspectives include related individual careers in ed- ucation, substance abuse, mental health, and employment. All the perspectives should contribute to the design of the initial sampling plan and ofthe basic data collection instruments. It is also impor- tant that the data that emerge from such a project be shared widely with the re- search community-especially those in- volved in the project design. Data col- lected from such a project shouIc! be fed into some public-use data facility quickly so that all participants and the research community would generally have access to it. One way to carry out the proposed longitudinal study would involve the cre- ation of a consortium of investigators to design the project and to be responsible for contracting with and overseeing an appropriate organization to collect the data. The instruments and procedures to be used would be negotiated within the consortium, ant! the results would be made available to the consortium and the research community generally for com- mon use and analysis, with appropriate safeguards to protect the subjects. This is the kind of approach follower] by the 201 Bureau of Labor Statistics in its National Longitudinal Stucly of Labor Force Par- ticipation, by the U.S. Department of Ed- ucation's High School and Beyond study of 1972 and 1980 high school graduates, and by the University of Michigan's Sur- vey of Income Dynamics. A consortium could also solicit other investigators to suggest data that should be collected or questions that should be addressed, which-would be incorporated to the ex- tent feasible and appropriate. This aspect is similar in concept to the use of a space shuttle to carry a diversity of experiments for individual investigators. The design and execution of the basic data collection instruments are analogous to the large overhead associated with creating and flying the space shuttle, which can gener- ally have capacity for augmenting its "payload" to satisfy a broad constituency of potential users. A very different way to carry out the study would be to identify an inclividual investigator skilled ant! knowledgeable in the relevant research issues and award that investigator a grant to organize what- ever staff, consultants, and advisers are necessary for the project. Such an ap- proach would undoubtedly attract a skill- fu! investigator, but any single researcher is likely to be limited to a particular the- oretical perspective and consequently to fad! to adequately address contributions from other perspectives. Any individual principal investigator is likely also to want to monopolize the data collected until all the results are assembled. A variant to this approach would be to involve multiple principal investigators at their own institutions, with a steering committee of the principal investigators and other researchers who would bring additional theoretical perspectives and methodological skills. With this ap- proach, each study would presumably be of smaller scale than the single one, and each investigator would be free to pursue
202 his or her design strategy, but the aggre- gation of investigators would represent a broad range of theoretical perspectives. CRIMINAL CAREERS AND CAREER CRIMINALS minants of termination of criminal ca- reers, possible policy choices will be sug- gested. The question of whether policy Such a program represents a compromise between the total commitment to a single investigator with its limited perspective and the coordination problems associated with a consortium. The panel has not tried to decide which approach to the design and implementa- tion of a major longitudinal stucly on crim- inal careers would be best a consortium of investigators in institutional setting, a single major project, or an aggregation of individual projects connected by a broad steering committee" but it does believe strongly that whatever approach is cho- sen should be structured so that a range of disciplinary perspectives can be reflected in the research program and that the data that are collected are shared with a broad range of researchers. The specific choice of organizational form should be decided by researchers and probable funding sources, such as the National Institute of Justice, the Bureau of Justice Statistics, the Office of Juvenile Justice and Delin- quency Prevention, the National Institute of Mental Health, the National Institute for Drug and Alcohol Abuse, and private foundations. SPECIFIC RESEARCH QUESTIONS Research on Intervention Strategies Much of the research on criminal ca- reers that the panel has considered is immediately applicable to the realm of policy choices, and a number of research efforts related to intervention strategies are warranted: one class of efforts relates to programs for successful prevention; an- other involves other interventions that may encourage career modification, inter- ruption, or termination. As research identifies dete~-~inants of participation in criminal careers or deter manipuiation of a statistically identified determinant will result in a change in the criminal behavior must then be explored through well-designed experiments. An important issue in this effort is the timing of an intervention and the tradeoffs in intervention effectiveness associated with earlier or later action: errors associ- ated with misidentifying high-risk popu- lations diminish with older ages, but later interventions may be less effective be- cause the offending pattern is more dif- ficult to change. Preventive Interventions The most attractive possibilities in pre- ventive interventions appear to be par- enting education, social learning and behavior modification programs, and ed- ~cational enrichment. Replications of an evaluation of a Head Start educational program in Michigan are clearly war- ranted`. Follow-up data on other Head Start experiments should be collected to compare the later criminal behavior of treatment and control groups. If those results appear encouraging, some new pilot experiments could be undertaken, leading eventually to larger-scale experi- ments. Analysis of the results from promising interventions requires careful attention to intermediate and long-term follow-up data to evaluate their effectiveness. Many preventive interventions appear to re- duce the frequency of conduct disorders or school problems in the short term (1 to 2 years) but are not successful in prevent- ing delinquency. The multitude of one- shot, nondefinitive research projects in this area suggests that another approach is warranted. The search for effective inter- ventions requires systematic, cumulative research with longer-term randomized
AN AGENDA FOR FUTURE RESEARCH experiments that are grounded in promis- ing theoretical orientations. Career Modification Strategies Other interventions are more appropri- ate after criminal activity has begun. Ca- reer mollification strategies would benefit from research on late teenage or early aclult offenders who spontaneously desist from offending and the identification of offenders who are most likely to respond to the intervention. Certain life events, such as employment, may encourage ca- reer termination, and this relationship could be translated into a program of job counseling for high-risk youth. Research on career-modifying interventions for younger adolescents has found behav- ioral mollification strategies and programs designed to enhance one's sense of re- sponsibility to be tentatively encourag- ing. The effectiveness of interventions directed at modifying the criminal career either through termination or reduction in seriousness or frequency is easily tester! through experiments. Such re- search efforts would benefit from a 3- to 5-year follow-up, possibly in conjunction with the major longitudinal effort recom- mended earlier. Last, postprison adjust- ment interventions that have not been entirely successful might be made more effective with a slight change in the pro- gram design to alleviate the work disin- centive effect that financial assistance may induce during the transition back to the community: for example, programs might require some community service in exchange for the assistance. Selective Incapacitation Decisions in the criminal justice sys- tem are currently based in part on an explicit intervention strategy of incapaci- tation. Many criminal justice practitioners already use fragmentary information 203 about criminal careers to make decisions about incarceration, which temporarily interrupts a criminal career. Some deci- sion makers may also make erroneous decisions by invoking wrong predictors that they mistakenly believe to be valicI. It is particularly important, therefore, to measure the effectiveness of incapacita- tion associated with current decisions that do try to identify offenders with higher individual crime rates and more serious offending patterns. This effect might be estimated from current self- report data by measuring to what extent higher-rate offenders are already receiv- ing longer sentences. The greater the se- lection of high-rate offenders in current practice, the less the benefit one can an- ticipate from cleveloping improver] pre- dictors. One can obtain some degree of selec- tive incapacitation in a number of ways without necessarily invoking any explicit predictor variables. If some particular crime types are associated with high-rate offending, then incapacitation strategies might be more effective by increasing the likelihood that offenders convicted of that crime type receive a sentence. This type of incapacitation policy invokes informa- tion that uniformly is accepted as appro- priatc the crime type of conviction. Re- search should focus on the benefit to be attained by such a policy. If only a small increment of predictability is provicled by using other more controversial predictors (such as employment status), then one can avoid the ethical problems posed by the use of those variables. There is a need for further validation ot the assumptions that underlie available estimates of the crime control and prison population effects of selective incapacita- tion policies and for analysis of the sensi- tivity of the estimates to violations of those assumptions. Information is espe- cially weak about crime spurts that occur just before arrest or just after release from
204 jai! or prison, about possible interactions between the frequency of offending (A) and the probability of arrest following a given offense (qY, about trends in q as offenders age, about the degree to which the crimes removed by incarcerating an offender are replaced by other offenders still in the community, ant! about the influence of selective incapacitation strat- egies on general deterrence effects. With such infonnation incorporated into more complete models, better estimates can be developed of the incapacitation effects of alternative policies on pretrial release, . . r assigning priorities tor court processing, sentencing, and parole release. Individual Deterrent Elects Another important set of interventions are the sanctioning actions ofthe criminal justice system. In particular, it is still uncertain whether time spent in prison has an individual deterrent effect on some offenders, in terms of reducing their postincarceration criminal activity, or a criminogenic effect, in terms of increas- ing individual crime rates and lengthen- ing criminal careers. Alternatively, both effects may occur, but on iclentifiably dif- ferent groups of offenders; if so, we need to learn how to distinguish the two groups. Measurement of Criminal Career Dimensions and Their Correlates It is clear that improved measures of criminal career dimensions are needed especially career length, participation in serious crimes, and the individual fre- quency of criminal activity for the gen- eral population and for high-risk subsets of the population. But perhaps more im- portant is the identification of particular behaviors, life events (such as marriage or first employment), or previous career characteristics (such as juvenile criminal CRIMINAL CAREERS AND CAREER CRIMINALS activity) that can be linked to the course of individual criminal careers. Such iclen- tification may require frequent Tongitudi- nal observations during key periods. Es- tablishing such linkages is particularly useful in view of the considerable diver- sity among offenders in their individual crime rates and their career length. If the individuals with the highest values of A and TR become candidates for selective interventions, then it is essential to de- velop information that will indicate the feasibility of such decisions, ant] how best to identify offenders who have the highest values. An important task in the study of crim- inal careers is linking empirical relation- ships to theoretical frameworks in order to avoid accumulating diverse indepen- dent correlations among fragmented sets of variables. The nature of the criminal career paradigm invites the application of dynamic theories to account for the clevel- opmental aspects of careers. Different theories might well be needed to explain the initiation of clelinquent behavior, the persistence of criminal activity into adult- hood, and the eventual termination of offending. It is likely that theoretical per- spectives from diverse disciplines can make important contributions to this ef- fort. Individual Crime Rates One of the most interesting and impor- tant issues in criminal career research has been the recent attention to developing estimates of individual crime rates, A, and the accumulation of reasonably consis- tent, independent estimates of those rates through both self-reports and official rec- ords. If results continue to be consistent, researchers will be encourged to under- take furler measurements in both mocles. Certainly furler measurement and analysis are needecl, especially for the serious crime types. In studying A,
AN AGENDA FOR FUTURE RESEARCH particular attention should be given to its longitudinal variation during an individ- ual's career. In what ways is the simoli- fied mode! of a stable inclividual crime rate in error, either in terms of age trends in the value of A while active or in shifts between high-rate and Tow-rate periods? Some researchers have iclentified the lat- ter scenario as particularly common among heavy narcotics users. Are there other life events that lead to a switch into or out of a high-rate period, and does the occurrence of high-rate periods follow identifiable patterns? The importance of individual crime rates is emphasized by the consiclerable skewness of the A distribution demon- strated in the Rand self-report inmate survey, which also highlights the noes! for attention on the correlates of A as a high research priority. Among the strong can- didates as determinants are characteris- tics of the juvenile career, particularly information about the age of first juvenile offense and specific offense types com- mitted, and this research will require con- tinuity of data in the juvenile and adult periods. There are some indications that the careers of adults in their mid-20s are likely to be more serious if the juvenile careers of those offenders started very early. Moreover, certain types or combi- nations of juvenile criminal behavior or its frequency may be important indicators of probable serious adult offending. These issues neecl further development and testing, and they probably could best be addressed in a longitudinal context. The influence of drug use on criminal careers, particularly inctiviclual crime rates, is a particularly complex issue. Se 1 rebus arug use as a juvenile or as an adult is indicative of a high rate of offending for those who engage in precatory crimes, such as robbery and burglary. If an effec- tive drug treatment program were avail- able to deal with the offender's depen- dency and if that depenclency couIct then 205 be reducer! along with the rate of of- fending, then cirug-related criminal be- havior could be modified by addressing the offender's drug problem. This type of intervention strategy requires identifying other indicators that will suggest which drug users are likely to be amenable to treatment. But the drug-crime relation- ship may also be closely tied to broader environmental factors, such as historical changes in the U.S drug trade, which may have contributed to the rise in aggregate crime rates in the 1970s. Thus, both the predictive features and the manipulable features of the drug-crime relationship warrant consiclerably more exploration, particularly in the context of available drug treatments and their effectiveness and the impact of environmental influ- ences. It is also important to study whether the information contained in an individu- al's history of arrests- the number, the mix of types, their rate of accumulation has more predictive power than the infor- mation in conviction records alone. Re- search on this question could contribute to the debate over using complete arrest histories rather than only conviction rec- ords. A part of this question should in- clude a comparison of adult only and combined adult and juvenile records to estimate the incremental value of infor- mation about juvenile careers in identify- ing serious adult careers. During the early stages of an adult career, full and ready access to the juvenile record is important; at present, there is often sharp discontinuity in records at the age of tran- sition (18 years in most jurisdictions). Furler research on the impact of this bifurcated court system on young adult offenders is neecled. Group Owning Exploring the effect of group influ- ences on criminal behavior might benefit
206 from the perspective of the criminal ca- reer paradigm. It would be desirable to identify groups of offenders whose crim- inal careers are linked through joint par- ticipation in some of the same offenses. Studies oftheir respective career patterns should indicate whether certain group patterns, such as a diversity of partners, are associated with high- or low-rate of- fending. It would be particularly impor- tant to determine to what extent some iclentifiable individual offenders serve as recruiters to offending and to what extent behavior can be explained as compliant responsiveness to recruitment efforts. The importance of group influences at initiation, the transition from group to solo offending, and the role of groups in later adult careers are other research is- sues related to group influences on crim inal careers. Career Length Initial research on career length has yielded some valuable insights regarding its relationship to age: termination rates during the 30s tend to be the lowest, a reflection of the fact that offenders who are still active in their 30s, especially those who started before age 20, are among the most committed to a criminal career. The use of existing data on the correlates of recidivism to explore pat- terns across follow-up periods of different lengths would help isolate other corre- lates of career length. Nonrecidivism can be a reflection of a combination of termi- nation and diminished A, and those two effects should be distinguished. The longer the follow-up period, the more likely it is that nonrecidivism reflects ca- reer termination. Long-range studies of career length will require individual ar- rest histories or self-report records for years after many careers have terminated. Two variables often associated with nonrecidivism are steady employment CRIMINAL CAREERS AND CAREER CRIMINALS and marriage: they may well be associ- ated with career termination, and they warrant closer study. The further ques- tion of whether explicit intervention through facilitating these transitions (e.g., through special counseling efforts) also has an effect should also be pursued. It may be that selection differences associ- ated with these transitions by offenders result in diminished criminality, but that interventions oriented at achieving those ends do not. This possibility can be tested in appropriately structured experiments. Last, ethnographic studies that focus in detail on individual life histories can be expected to provide a richness of detail about factors contributing to termination of criminal careers that would not be available in even the most careful survey study. Participation Future research on participation is likely to benefit less from further mea- surement and more from attention to im- portant correlates of initiation of criminal careers and the different correlates of mi- nor and serious criminal involvement. In particular, more research is needed on life events that may be linked to criminal behavior in offender populations but that have not been adequately stuclied in a longitudinal context. Examples of these include abusive family situations, paren- tal alcohol or drug abuse, divorce or death of a parent, or other traumatic life events, particularly if they occur in the preschool or elementary school years. This type of research is particularly valuable for indicating which indivicluals are most likely to initiate delinquent ca- reers, which individuals are likely to commit only minor offenses and then ter- minate their delinquent activity, and which individuals are likely to move on to more serious delinquency. If such indica- tors could be clevelopecl, the potentially
AN AGENDA FOR FUTURE RESEARCH effective intervention strategies clis- cussec! earlier might be more effectively targeted. New Directions in Measurement and Modeling Much too often the two approaches to · · · 1 · measuring criminal career c .lmenslons- self-reports and official records have been presented as mutually competitive. They are much more appropriately viewer] as complementary mocles of ob- servation of a common unclerlying proc- ess, ant! specific effort shouIc! be clirectec! at deriving estimates, especially of A ant! its distribution, that use information from both ciata sources. The strengths of each approach will help to illuminate, ant! ul- timately to correct, the sources of error in the other. Specifically, it is important to learn more about some key linkages, particu- larly that between the arrest process that official records depict anc! the unclerlying crime process that is of primary concern. A key element in this linkage is the sam- pling probability from crimes to arrests- the probability of an arrest conclitional upon a crime en c! any bias it contains or correlations it might clisplay with the un- clerlying incliviclual crime rate. Self- reports, in conjunction with official rec- ords for the same inclivicluals, provide an opportunity for exploring this linkage. Such exploration is an issue of high pri- ority, largely because improver] insight into the error structure of the two ap- proaches of measurement, the connection between their error structures, and the link between arrests and the unclerlying crime process wouIc! increase the useful- ness of the large numbers of official- record ciata sources that are available as primary records for studying criminal ca reers. Most research on trends in crime seri- ousness over a criminal career has re 207 veaTec! incliviclual diversity, albeit with some greater tendency for offenders to repeat the previous offense type or to stay within the group of property or violent offenses. Most of those results, however, derive from observations of the arrest process, and inferences about patterns in the underlying crime process have proven to be rather difficult. This area is of considerable interest and importance for developing knowledge about the structure of criminal careers. It also has import for policy because of the signifi- cant influence that crime-mix patterns could have on incapacitative effects. One important stumbling block has been the fact that most of the data on crime-mix patterns have reflected the arrest process, and arrest patterns are not necessarily reflective ofthe underlying crime process because of the great variability in arrest probability (q) across different crime types. The arrest process has been domi- nant because the sequence of arrests is well recorded in official arrest histories, and the sequence of crimes is rarely asked for in self-reports. These issues could be pursued more directly for serious offenses in self-report questionnaires that probe into sequences of participation by first asking respon- dents if they have ever done a particular offense and then asking them when they first did it and when they last did it. The r issue ot c ranges in crime mix may pOSSl- bly be more effectively addressed by focusing on changes in crime-type partic- ipation patterns rather than the micro- analysis of the sequence of crime types that is easily available from arrest records. Further research on estimates of incapacitative effects is also needed. Es- timation of the incapacitative effects of imprisonment are most commonly com- puted using a formula derived from a mode! first introduced in 1973. This model uses standard assumptions for trac- tability: homogeneous offenders, a Pois
208 son stream of crimes, exponential career lengths, parameters that are constant over time, and prison sentences that are short compared with career lengths and incle- pendent of prior record. The panel's anal- yses have shown that the incapacitation effects estimated by Greenwood (1982) are very sensitive to the effects of finite career lengths. Further exploration is needed on the sensitivity of the results to mode! assumptions, and changes in the mode] are needed to reflect the conse- quences of the considerable heterogene- ity in values of A. Also, the mode! should take account of the fact that sentences often do reflect predictive considerations and the possibility that sentences could be related to a judge's estimate of an individuaT's offending frequency. Recent developments in research on stochastic processes associated with hier- archical distributions offer promise for further development of the incapacitation model (Lehoczky, Volume II). Any such model clevelopment should be tied closely to knowledge about the structure of criminal careers and to the potential incapacitative effects associated with those criminal career histories. Also, models should be extender] to reflect var- ious theoretical perspectives on the fac- tors that influence criminal career climen- sions. One such model, reflecting a labor-market viewpoint, was clevelopec! by Flinn (Volume II); other theoretical approaches should be pursued. In exploring various selective incapac- itation polices, it would be desirable to measure the selective incapacitation ef- fects associated with several classification instruments. Any such attempt should en- gage in simulated experiments using in- dividual arrest data. It is important in the simulated experiments that the data set on which the experiments are run be different from the one that gave rise to the classification rule in order to avoid shrinkage problems (see Chapter 61. CRIMINAL CAREERS AND CAREER CRIMINALS Seemingly powerful predictors that emerge in Me first stage can then be tested experimentally for their potential usefulness in actual applications. SUPPORTIVE ORGANIZATIONAL ARRANGEMENTS In this final section we identify two organizational arrangements that could facilitate the further development of a cohesive research community and the ex- pansion of knowleclge about criminal ca- reers: continuation of the current Crime Control Theory Program at the National Institute of Justice (NIl) and develop- ment of a criminal career research data repository. Continuation and Expansion of the Crime Control Theory Program Much of the research on criminal reers reviewed by this pane! has been supported by the Crime Control Theory Program at NIT, which also supported the work of the panel. Over a period of 6 years, with a limiter! budget of less than $l million per year, this program has made some important contributions to knowledge about criminal careers and their policy usefulness. We believe that much of the success of this program has been a direct consequence of its research management strategy. The program man- agers provide general guidance on the policy objectives; give heavy emphasis to peer review for research proposals; dis- play a strong commitment to the continu- ity of the research stream, to the clevelop- ment of a research community, and to interchange of information through an- nual meetings at which research results are critically examined by peers and col- leagues; and Took to the scientific review process for filtering valid from invaTicl results. This strategy, which the panel endorses, has long been associates! with
AN AGENDA FOR FUTURE RESEARCH effective progress in medical research and in the physical sciences. The panel hopes that this effort will not only be continued, but that additional resources will be provider] to permit the program to support some of the research questions discussed in this chapter. Criminal Career Data Repository A major need common to all research on criminal careers is access to a rich set of data on individuals. Ideally, such data would record cletailed information on how the sample was drawn; on crime and arrest sequences; on each inclividuaT's early childhood experiences; on his par- ents, siblings, and peers; on school expe- riences ant] work experiences; on deviant behavior of various sorts; and on interac- tion with the criminal justice system. No single data source yet collected contains so rich a set of information on an appro- priately broact sample. In some cases, data are available only for juvenile years; in other cases, only aclult arrest events are available; and in only a few cases are indiviclual background data available. To- gether, however, the wide variety of in- formation that has aIreacly been collected and used in indiviclual criminal career research projects represents a rich collec- tion on which consiclerable secondary analysis is warrantee! and should be pur- sued. In order to facilitate such secondary analysis, it would be appropriate to orga- nize and code those data sets so that they can be made accessible to a large commu- nity of potential users for comparative research and cross-vaTidation. The com- munity working with the data would be in a strong position to test prediction moclels developed in one setting or from one sampling frame against data collected 209 in another and to cross-validate various stuclies. Even within a single study, the data from the initial construction sample from which a particular predictive mode! was developed could be tested against a validation sample, which would develop measures of shrinkage as part of valida- tion. The initial data sets to be maintained in a criminal career data repository should include, at a minimum, the Rand inmate surveys, the National Youth Survey of Elliott et al., the official-record arrest data collected ant! analyzed by Blumstein and Cohen, the clata on two Philaclelphia co- horts collected by Wolfgang et al., the various cohort data of Shannon, Robins, McCorcI, and the London longitudinal data of West and Farrington. The location of such a repository would have to be explored, but one possibility is the al- ready-established Criminal Justice Ar- chive an(1 Information Network (CIAIN) at the University of Michigan, the current repository for all data collected with NIT support; the data maintainer] in the C.lAIN are available to any user. The field of criminal career research is beginning to yield valuable information on the dimensions of criminal careers- participation, frequency, seriousness, and career length and the factors associated with each dimension. Advances in mea- surement an(l modeling have been instru- mental in the accumulation of knowledge about individual careers. Additional re- search that a(ldresses the questions pose in this chapter will contribute to more effective strategies to prevent, mollify, interrupt, or terminate criminal careers. Progress can be expected from secondary analysis of available data, from well- designed experiments, and, most impor- tantly, from a major program of Tongitudi- nal study.
